A recent study conducted by Stanford Medicine reveals that artificial intelligence (AI) can enhance intravenous nutrition (IV nutrition) for premature babies, which has the potential to reduce medical errors, save time and costs, and improve care, especially in low-resource settings. This study, which will be published in Nature Medicine on March 25, is one of the first to demonstrate how an AI algorithm can assist doctors in making better clinical decisions for sick newborns.
The AI tool utilizes information from preemies’ electronic medical records to predict the exact nutrients and quantities they need. It was trained on data from nearly 80,000 past prescriptions for IV nutrition, linked to the health outcomes of these tiny patients. The goal is to optimize IV nutrition, known as total parenteral nutrition (TPN), for premature babies who are unable to absorb nutrients through their digestive systems.
Nima Aghaeepour, PhD, the senior study author and associate professor at Stanford, emphasizes that IV nutrition is crucial for preemies, but it is also one of the most error-prone processes in neonatal intensive care units (NICUs). Every day, neonatologists, pharmacists, and dietitians manually create TPN prescriptions for each premature baby, which can lead to inconsistencies and errors.
Challenges of IV Nutrition for Preemies
Around 10% of babies are born prematurely, and those born more than eight weeks early require IV nutrition to survive. The current prescription process involves multiple experts and takes several hours. This makes it a high-risk and time-consuming task. Moreover, there’s no blood test to directly measure whether a preemie is receiving the right amount of calories or nutrients each day.
“We can’t approximate what the placenta is doing,” said David Stevenson, MD, a co-author of the study. “Nutrition is one of the areas of neonatal intensive care where we are weakest.”
Currently, each prescription is individually created based on the baby’s weight, stage of development, and lab results. However, this process often leads to errors, which could negatively affect the baby’s health, making it essential to streamline and standardize the process.
AI Algorithm and Standardized Nutrient Formulas
In the study, the AI algorithm was trained on a decade of data from Stanford’s Lucile Packard Children’s Hospital, including 79,790 prescriptions for 5,913 premature patients. The researchers sought to determine if AI could create standardized formulas for IV nutrition, which could be adjusted daily as the baby’s condition changes.
Using the trained algorithm, the researchers created 15 distinct formulas for IV nutrition. These formulas were designed to meet the nutritional needs of a wide range of premature babies, based on the patterns found in the data. The researchers tested the AI-generated prescriptions against real prescriptions and found that doctors consistently preferred the AI-generated prescriptions.
The AI algorithm was also able to predict which of the 15 formulas each baby might need, adjusting the formula recommendation as the baby grew and their condition changed. This daily adjustment helps ensure the baby receives the right nutrients throughout their hospital stay.
Results and Future Steps
When the research team compared AI-generated prescriptions to actual prescriptions, they discovered that babies whose nutrition prescriptions were aligned with the AI recommendations had better outcomes. The risk of complications like sepsis, mortality, and bowel disease was lower for these babies.
The next step is to conduct a randomized clinical trial to test how babies receiving AI-generated prescriptions fare compared to those on manual prescriptions. The AI system will still require oversight by clinicians, but once the prescriptions are approved, standard nutrient formulas could be readily available, reducing the need for a large expert team and compounding pharmacy.
Impact on Global Healthcare
The researchers believe that AI could make IV nutrition more accessible, especially in low-resource settings. Standardizing formulas and reducing the need for expert teams could make it easier for hospitals in underfunded regions to care for premature babies. The technology could also reduce healthcare costs by making the process faster and more efficient.
“We hope AI will enhance medicine by making doctors better and more efficient, allowing them to spend more time with babies and families, providing comfort and reassurance,” said Stevenson. The AI technology, alongside human expertise, could provide top-notch care to more babies worldwide, particularly those in lower-income countries or regions with fewer medical resources.
This research represents a step forward in integrating AI into healthcare, improving outcomes for vulnerable patients, and making specialized care more accessible globally.
You Might Be Interested In:
-
Drinking Diet Drinks in Pregnancy Raises Overweight Risk in Children, Study Finds
-
Diets Linked to Healthy Aging: 30-Year Study of 100,000 People Highlights Key Findings
